undefined

In-memory clusters and eliminating SAN for cost savings on 10B data sets a year

Case Study

XING is a social networking platform for business professionals, with 16 million members worldwide. Based in Hamburg, Germany, XING members organize and attend events, find job opportunities, and connect with other experts in 80,000 specialist groups and 16 languages. XING uses intelligent data analysis to target networking and recruiting information for its member companies that advertise directly on the platform.

The Facts

media-entertainment
Germany
2327
Cloud

The Gist

The challenge

As XING grew, its legacy database reached a limit, with data processing often exceeding 24 hours. They needed a scalable solution for complex analysis of 10 billion data sets a year, clustering for better speed, and lower administrative costs.

The solution

In just four weeks following a benchmarking test, XING integrated Exasol into its existing infrastructure. ETL (Extract Transform Load) processes were converted consecutively into the new solution.

“Queries that were previously unthinkable are now possible.”

Dr Ulrich Fricke, Data Warehouse and Reporting Team Leader

Benefits

Benefits for Xing

Within two years of migrating to Exasol’s data warehouse, XING’s data volume increased 5x. With improved performance, what were previously impossible evaluations of data sets are now commonplace.

XING pays for what it needs.

Exasol’s integrated distributed data storage eliminates the need for a storage area network (SAN). Cluster compression algorithms enable efficient main memory, keeping system costs low.

Self-learning adjusts to user preferences.

Indices are automatically laid out, data is compressed, and SQL instructions from the query optimiser are tuned. XING can process millions of queries per month of the parallelised and complex loading and transformation processes, as well as interactive user queries.

In-memory cluster technology

and compression algorithms allow XING to process and analyze large volumes of data in seconds. Query processing takes place in the RAM, eliminating slow access to hard drives, pre-calculation, and aggregation.

Columnar storage.

Multiple nodes contain two processors each with the individual servers connected in a cluster, where data is automatically divided.

Start your 30-day free trial